PII proteins are one of the most widely distributed signal transduction proteins in Nature, being ubiquitous in bacteria, archaea and plants. They act by protein–protein interaction to control the activities of a wide range of enzymes, transcription factors and transport proteins, the great majority of which are involved in cellular nitrogen metabolism. The regulatory activities of PII proteins are mediated through their ability to bind the key effector metabolites 2-OG (2-oxoglutarate), ATP and ADP. However, the molecular basis of these regulatory effects remains unclear. Recent advances in the solution of the crystal structures of PII proteins complexed with some of their target proteins, as well as the identification of the ATP/ADP- and 2-OG-binding sites, have improved our understanding of their mode of action. In all of the complex structures solved to date, the flexible T-loops of PII facilitate interaction with the target protein. The effector molecules appear to play a key role in modulating the conformation of the T-loops and thereby regulating the interactions between PII and its targets.
The PII signal transduction proteins were first identified in 1969 during studies of post-translational modification of the activity of Escherichia coli GS (glutamine synthetase) . At that time, two protein fractions, PI and PII, were described: PI subsequently being determined to contain both the adenylyltransferase (GlnE) responsible for adenylylation of GS, and a uridylyltransferase (GlnD) responsible for the reversible uridylylation of PII. PII regulated the activity of GlnE and was later shown to be encoded by glnB. PII proteins are now recognized as one of the most widely distributed signal transduction proteins in Nature, being ubiquitous in bacteria, archaea and plants . Many bacteria and archaea encode multiple PII proteins, whereas only a single copy is present in the cyanobacteria and in plants .
PII proteins function by protein–protein interaction, whereby they have been shown to control the activities of a wide range of targets, including enzymes, transcription factors and membrane transport proteins. These targets are almost invariably involved in cellular nitrogen metabolism, and hence PII proteins are considered to be pivotal regulators of nitrogen metabolism in most prokaryotes. More recently, it has also been proposed that their functions may extend to the co-ordination of nitrogen and carbon metabolism and to sensing of cellular energy status . The key to the sensory properties of the PII proteins is their ability to bind the effector molecules 2-OG (2-oxoglutarate) and ATP or ADP. In the present review, we discuss recent advances in our understanding of this process.
The structures of PII proteins
PII proteins are homotrimers with 12–13 kDa subunits that display a highly conserved structure. The trimer is a compact cylindrically shaped molecule from which three long exposed loops (the T-loops) protrude (Figure 1A). The T-loops are significantly conserved in sequence, but are structurally very flexible and can adopt various conformations. They are vital for PII interactions with many of their targets and are also often sites of reversible covalent modification [4,5]. In addition to the T-loops, PII proteins are also characterized by three lateral inter-subunit clefts, within which are two smaller loops (the B- and C-loops) (Figure 1). PII proteins have been divided into three subfamilies, encoded by the genes glnB, glnK and nifI . GlnB and GlnK are closely related and are often encoded in the same organism. A considerable number of X-ray structures have been solved for both GlnB and GlnK. NifI proteins are more distantly related: they occur in nitrogen-fixing archaea and function in regulating nitrogen fixation ; however, no structures of NifI proteins have been solved. Recently, a detailed phylogenetic analysis identified a new group of PII proteins, designated PII-NG . Very little is known about these proteins, but their structural genes are frequently linked to genes for heavy metal transporters and it is deduced that they may be involved in metal homoeostasis. We will not discuss this group further in the present paper.
The structure of the E. coli PII protein GlnK
The interactions of effectors with PII proteins
Early studies with Escherichia coli GlnB implicated ATP, 2-OG and Mg2+ as potential effector molecules that facilitate PII-mediated regulation. It is now apparent, from studies in a number of systems, that ATP and 2-OG bind to PII at micromolar concentrations and their binding is synergistic [5,7,8]. However, as ATP binds avidly in the presence of 2-OG, it was considered unlikely that the nucleotide plays a regulatory role in vivo. It was only recognized relatively recently that PII proteins could also bind ADP and that this might be physiologically significant [3,9,10]. Biochemical studies on E. coli GlnB and GlnK showed that ADP acted antagonistically to 2-OG [3,11]. The ability of PII proteins to bind ATP or ADP alternately substantiated earlier suggestions from studies in Rhodospirillum rubrum that, in addition to sensing 2-OG levels, PII proteins might have the capacity to sense adenylate energy charge [3,9].
The site of ATP binding was determined from the crystal structure of E. coli GlnK co-crystallized with ATP . The nucleotide binds in the lateral clefts between the subunits, where the B-loop from one subunit and C-loop from another contribute to a nucleotide-binding pocket (Figure 1B). The B-loop includes a consensus sequence (G84XXGXGK) identical with that found in other nucleotide-binding proteins and referred to as the Walker A motif. Key contacts to the phosphate groups include Gly87 and Gly89 in the B-loop and Arg101 and Arg103 from the C-loop, all of which are highly conserved residues in the PII family. When ADP was identified in the structure of the complex between E. coli GlnK and the ammonia channel AmtB (see below), it occupied the same site as ATP (Figure 2A) and made very similar interactions, the β-phosphate of ADP occupying a very similar position to that of the γ-phosphate in the GlnK–ATP structure. ADP has also been identified in structures of PII proteins from Methanococcus janaschii (PDB code 2J9D) and Thermotoga maritima (PDB code 1O51), but the physiological significance of these structures is unclear.
Binding of effectors to PII proteins and the influence of 2-OG
The 2-OG-binding site has proved much more elusive. In a structure of M. janaschii GlnK1, 2-OG was found to be bound to one of the T-loops of the trimer . A recent structure of Azospirillum brasilense GlnZ (a GlnK homologue) shows a more convincing location for 2-OG, bound near to Mg-ATP within the lateral cleft (Figure 2C) . This structure fully rationalizes the synergistic binding of 2-OG and ATP and is consistent with data from E. coli GlnB showing that 2-OG binds even when the T-loops are deleted . The GlnZ structure now offers a basis on which to understand the role of effector binding, namely in influencing the conformation of the T-loops. Two key residues in this structure are the highly conserved Gln39 and Lys58, each of which lie at the base of the T-loop (Figure 2C). Both ATP and 2-OG participate in the co-ordination of the Mg2+ ion, three bonds provided by the phosphates of ATP, two by the 2-oxo acid moiety of 2-OG and the final one by the Gln39 side chain. 2-OG also forms a salt bridge with Lys58 (Figure 2C). The significance of these interactions becomes apparent from recent structures of PII proteins complexed with some of their targets (see below). However, it should also be noted that, whereas the roles of 2-OG and ATP/ADP as PII effectors has been considered to be a universal feature of these proteins, Bacillus subtilis GlnK has been reported to bind 2-OG only weakly  and Archaeoglobus fulgidus GlnK2 shows no 2-OG binding .
PII protein targets
The initial studies of PII were in the context of regulation of GS activity and specifically the regulation of adenylyltransferase activity by GlnB. Since then PII proteins have been found to regulate a wide variety of targets including enzymes, transcription factors and membrane transport proteins (see [4,6,18] for reviews). Bioinformatic analysis of the likely evolution of PII proteins supports the concept that the ancestral PII evolved in association with the ammonia channel AmtB [2,19]. The key role of GlnK proteins in forming a complex with the trimeric AmtB to regulate ammonia influx into cells rationalizes the trimeric nature of PII proteins. However, it is clear that the regulatory properties of PII proteins have subsequently been adopted to co-ordinate regulation of a whole range of cellular activities [4,18]. This means that PII proteins are able to interact with a wide variety of different protein structures and that the mechanisms of these interactions are likely to be equally varied. Of particular interest in this regard is the fact that many of the PII targets are not trimeric, raising questions about the likely stoichiometry of many of these interactions. A detailed understanding of the molecular mechanisms whereby PII proteins control the activities of their targets depends on structural information on the specific complexes combined with detailed information on the physiological conditions under which complex formation occurs. To date such information is relatively limited. However, the solutions of the AmtB–GlnK , PII–NAGK (N-acetyl-L-glutamate kinase)  and PII–PipX  structures offer new insights into the properties of PII proteins that make them so adaptable for interaction with multiple targets.
Structures of PII proteins complexed with targets
In nearly all prokaryotes, the amtB gene is co-transcribed with glnK, strongly suggesting that these two proteins function together . Evidence for this interaction was obtained initially from studies in E. coli [20–22] and the structure of the complex was solved in 2007 [10,23]. In previous structures of the isolated PII protein, the T-loop was often unresolved, reflecting its flexible nature, or if it was resolved, then the observed conformation was constrained by crystal contacts and hence did not necessarily reflect a structure that the T-loop might adopt naturally. In the AmtB–GlnK complex, the T-loops of GlnK adopt a previously undescribed extended structure with the conserved residue Arg47 at the loop apex. Each of the T-loops inserts deeply into the cytoplasmic pore exit of the AmtB conduction channel, thereby blocking ammonia conduction (Figure 3A). Another key feature of the complex was the presence of ADP in the nucleotide-binding site of GlnK. No nucleotide had been added to the crystallization and the complex had been purified intact directly from ammonium-shocked E. coli cells. Hence it was concluded that the presence of ADP was likely to reflect a physiologically relevant state, and that ADP might play an important role in influencing the T-loop structure and promoting complex formation [10,23].
Different modes in which PII can interact with its targets
In recent studies, we have monitored in vivo levels of PII effectors in E. coli during association and dissociation of the AmtB–GlnK complex. When GlnK is not complexed to AmtB, intracellular levels of 2-OG and ATP are relatively high . Consequently, the structure of A. brasilense GlnZ with bound 2-OG and Mg-ATP (Figure 2C) is probably a reasonable model for the status of cytoplasmic E. coli GlnK. When cells become nitrogen-sufficient, intracellular 2-OG levels fall significantly and there is a transient rise in the ADP pool . Furthermore, in vitro studies show that binding of ADP to GlnK is critical for complex formation with AmtB. Comparison of the structure of the GlnK nucleotide-binding site in the AmtB–GlnK complex with that of GlnZ reveals that, in the absence of 2-OG, Gln39 relocates to form a new bond to Lys58 (Figure 2A), thereby influencing the conformational space occupied by the T-loop. Hence the switch from bound 2-OG and ATP to bound ADP is likely to be the key event in promoting AmtB–GlnK complex formation. In summary, the location of the 2-OG-binding site near the base of the T-loop offers a likely explanation as to how 2-OG sensing by PII proteins might be translated into alternative conformations of the T-loop.
The second structure of a PII complex solved recently is that of PII with the enzyme NAGK. This interaction occurs in both cyanobacteria and plants, and structures of the complex have been solved from both Synechoccocus elongatus and Arabidopsis thaliana [24,25]. NAGK is the controlling enzyme of arginine biosynthesis and, in these organisms, arginine is used as a nitrogen store, thus explaining a role for PII in controlling NAGK activity. NAGK is a hexamer made up of two back-to-back homotrimers and, in the complex, this is sandwiched between two PII trimers (Figure 3B). The only features in common with the AmtB–GlnK complex are the symmetry and 1:1 stoichiometry, together with the fact that the T-loops again constitute a major interface with the target. Arginine is a feedback inhibitor of NAGK activity, and the binding of PII rearranges both the allosteric arginine inhibition site and the catalytic centre of NAGK, thereby decreasing the affinity for arginine and increasing the affinity for the substrate NAG (N-acetylglutamate). No effector molecules were present in the S. elongatus complex structure. Remarkably, for the A. thaliana complex, well-diffracting crystals were only obtained in the presence of Mg2+, ADP, NAG and arginine, but nevertheless Mg-ATP was found to be bound to PII in the crystal. The authors conclude that it was likely to have been bound to the PII when it was purified from E. coli . In both PII–NAGK complexes, a salt bridge is present between a conserved lysine residue and glutamate residue at the base of the T-loop (Figure 2B). Hence, as in the AmtB–GlnK complex, 2-OG binding to PII will predictably alter the conformation of the T-loop.
The effects of effector molecules on the PII–NAGK complex have been analysed in a number of studies for both S. elongatus and A. thaliana [26–29]. For the cyanobacterial system 2-OG completely abolishes the interaction providing that Mg-ATP is also present [28,29]. The plant complex is much more resistant to 2-OG and only shows gradual inhibition of complex formation at 2-OG concentrations greater than 1 mM [26,29]. ATP promotes complex formation in both systems, but, whereas ADP is a potent inhibitor of complex formation in S. elongatus, it is not in A. thaliana . This is presumptively due to the very high affinity of AtPII for ATP such that it is not easily displaced by ADP [25,29]. Hence, it may be that the two systems have evolved somewhat different responses to cellular physiology, but further studies on fluctuations in effector pools in both systems and their correlation with NAGK activity would be very informative.
The third PII complex structure is that of S. elongatus PII with another of its targets PipX . PipX is a co-activator of the cyanobacterial transcription factor NtcA, and PII interaction with PipX antagonizes its ability to activate NtcA . The complex has a 3:3 stoichiometry, such that one PII trimer binds three molecules of PipX (Figure 3C). The PII T-loops are in an extended conformation resembling, but not identical with, that seen in the AmtB–GlnK complex and PipX is ‘caged’ between them. Although no effector molecules are bound in this structure, ADP is known to increase the affinity of PII for PipX . It is proposed that binding of 2-OG to PII antagonizes ADP binding and triggers conformational changes in the T-loops that facilitate the release of PipX . Such a proposal is entirely consistent with the role of NtcA in activating gene expression in response to nitrogen limitation. Hence, as S. elongatus becomes progressively nitrogen-starved, cellular 2-OG levels are predicted to rise. Binding of 2-OG and ATP to PII would modify the T-loop conformation, thereby ensuring that PipX is released from the interaction with PII and becomes free to activate NtcA.
Studies of PII proteins from a wide variety of organisms have confirmed that 2-OG, Mg2+, ATP and ADP are the key effector molecules that regulate the abilities of PII proteins to interact with their targets. Recent structures of PII proteins have identified the effector-binding sites and, together with multiple sequence alignments, these strongly suggest that the mode of binding for each of the effectors is conserved. It is now apparent that the interaction of 2-OG with PII is the key indicator of cellular nitrogen limitation, but that ATP and ADP also play an important role in regulating interactions between PII proteins and their targets [11,14]. However, the precise role of ATP and ADP as regulators of PII interactions requires further study, and the concept that PII proteins have the capacity to sense adenylate energy charge [3,9] has yet to be fully validated.
From the structural data available at present, it is clear that the T-loops play a key role in most interactions of PII proteins with their targets. Their conformational flexibility means that they are able to mediate the interaction with targets that themselves have widely different structures (Figure 3). We can also now begin to understand how effector binding can influence the T-loop conformation and hence the role that effectors have in the signal transduction process. However, it should also be noted that the apparent ability of PII proteins to mediate the formation of ternary protein complexes, in which the AmtB–GlnK complex recruits a third partner [18,32], suggests that not all PII interactions will necessarily involve the T-loops. To date, we only have one system, AmtB–GlnK, where we are approaching a holistic description of the PII-mediated signal transduction process. Consequently, there is a definite need for the solution of new structures of PII proteins bound to other targets and for more physiological studies that can link cellular metabolism to underlying changes in effector pools.
Enzymology and Ecology of the Nitrogen Cycle: A Biochemical Society Focused Meeting held at University of Birmingham, U.K., 15–17 September 2010. Organized and Edited by Jeff Cole (University of Birmingham, U.K.), Rosa María Martínez-Espinosa (University of Alicante, Spain), David Richardson (University of East Anglia, Norwich, U.K.) and Nick Watmough (University of East Anglia, Norwich, U.K.).
This work was supported by the Biotechnology and Biological Sciences Research Council (BBSRC) [grant number BB/E022308/1 (to M.M.)].